Bioenergetics and the Evolution of Cellular Traits
Most cellular traits are products of genes, so the cost of individual genes is a good starting point for an enumeration of cellular trait costs. Gene costs are associated with replication, transcription, and translation, and cost estimates for the full gene complement of a diverse set of organisms reveal that the average relative energy cost of individual genes decreases with cell volume and that translation costs are higher than the costs for transcription and replication (48). Average gene costs, relative to the cell budget, are shown for bacteria and eukaryotes in Figure 2b. Wagner (98), by a similar method, estimated that the relative cost of duplicating the median gene is 4.68 × 10–5 in Saccharomyces cerevisiae, which (given the cell volume) is consistent with the relative costs in Figure 2b.
Energetic costs of individual macromolecules can be computed when the monomer composition is known (1, 3, 48, 72, 74, 75, 90, 98). As examples, the energetic costs of the small protein calmodulin, the protein complexes ATP synthase and dynein-2, and the protein- and RNA-containing ribosome are shown in Figure 3. Cellular trait costs are sums of the costs of gene products, including, but not limited to, proteins, membranes, and cell wall components. Some cellular traits are composed of only proteins (e.g., the circadian clock, gas vesicles, and the cytoskeleton). The energetic costs of these (and other) traits are listed in Table 1.
Cellular traits such as vesicles, organelles, and (some) cellular appendages contain membranes, and an overview of membrane-associated costs is shown in Table 1 and Figure 3. For the membrane costs of mitochondria in ciliates, a power law relation has been estimated: 2.3×109Vcell0.98 ATP (50). If the exact composition of an organelle is not known, volume fractions, of the organelle compared to the volume of the whole cell, can be used as a (rough) proxy for cost. This, for example, allows for a comparison of mitochondrial and chloroplast costs, with mitochondria at volume fractions ranging from 0.46% to 15% (88) being less costly to cells than chloroplasts with volume fractions frequently between 30% and 40% (78, 96). These ranges reflect natural variation, not measurement error.
Flagella are ubiquitous cellular appendages used for swimming and their cost can be estimated from structural studies of the protein components and (when present) from the membrane area (51, 83) (Figure 3; Table 1). A survey of 200 species of bacteria and single-celled eukaryotes showed that in most cases the relative construction cost of flagella is between 0.1% and 40% (86) (Figure 2c). The relative costs of individual cilia (flagella) and cirri (cilia bundles) for many species of ciliates are shown in Figure 2d. These costs reveal what benefit the addition of a single cilium or cirrus must bring to the cell before the net fitness change is positive. They also reveal that the addition of even a single cilium or cirrus can be detected by natural selection (by comparison with the inverse of the effective population size, 1/Ne, see below) (Figure 2d).
The nucleus is a defining characteristic of eukaryotic cells, and the energetic costs of the envelope can be evaluated by examining its membranes and pore complexes. Any benefits proposed for the evolution of the nucleus (11, 36) need to exceed these costs for an evolutionary scenario to be plausible. Nuclear membrane cost can be estimated from nuclear surface area. For ciliates, this yields an energy cost of 6.8×109Vcell0.54 ATP (macronucleus) and 1.5×109Vcell0.41 ATP (micronucleus) (50, 86) (Figure 2e). For other species, including phototrophs and mammals, the scaling is similar: 7.8×109Vcell0.48 (52, 86). Using the nuclear volume as a proxy for the relative cost of the whole nucleus (including, e.g., the genomic content), the relative nuclear energetic cost for ciliates is given by 2.82Vcell−0.41 (macronucleus) and 0.16Vcell−0.54 (micronucleus) (50). For a typical ciliate cell volume of 105 μm3 this yields relative costs of 2.5% (macronucleus) and 0.032% (micronucleus). The nuclear volume fraction can be higher in multicellular organisms, with angiosperm cells devoting 20% of their volume to the nucleus (58). Algae have nuclear volume fractions between 4.4% and 20% (78). Nuclei are studded with NPCs that regulate the transport between the nuclear lumen and the cytoplasm. An individual yeast NPC costs 1.3 × 107 ATP, and for vertebrates this is 2.9 × 107 ATP, using respective molecular weights of 50 Mda (2) and 110 MDa (7) (Figure 3). When an average of these values is used, the cost of all NPCs combined scales with the nuclear volume as 7.0×108Vnuc0.78 ATP (52).
Energetic costs have also been studied for trichocyst discharge (Table 1), compatible solute synthesis and ion import in response to osmotic stress (69), vesicle release and action potentials (4), membrane protein insertion (30), membrane curvature generation (92), cellular sensing (29), cell cycle oscillator (85), noise regulation by microRNAs (33), and silicon use (76).
Natural selection is not all powerful. It can be counteracted by mutation pressure and genetic drift (46, 47). An extra base pair in a gene, an extra amino acid in a protein, and an increase in expression level all come with energetic costs that are detrimental to cell growth and survival, unless these changes come with additional positive effects on cellular functioning. Even in the absence of such positive effects, the modifications can still be fixed in the population if the energetic cost of an addition is small enough compared to mutation pressure or genetic drift. Gene frequencies in finite populations are subject to random fluctuations (genetic drift) with magnitudes proportional to Ne. Ne values are generally far smaller than those of census population sizes and depend on genetic phenomena such as recombination and background mutation rates as well as on demographic factors. Fitness effects smaller than (roughly) 1/Ne are invisible to selection (47). To determine the fate of a cellular adjustment, be it the accidental duplication of a gene or a single amino acid insertion, it is therefore important to compare the fitness effects to 1/Ne.
The energetic cost of the duplication of any gene in the S. cerevisiae genome, and its accompanying doubling of the expression level, has been estimated (98), showing that for all genes a duplication event is costly enough to be detected by natural selection. In the same organism, an expression-level change of >10% for mRNA and >2% for protein, for a gene with median expression level, would be visible to natural selection; below these thresholds changes could occur neutrally (98). Gene costs in E. coli conform to the conclusion that gene duplications are visible to selection, but gene costs in Caenorhabditis elegans and Arabidopsis thaliana are low enough, compared to the effects of genetic drift, that many genes can be duplicated without opposition from natural selection (48) (Figure 4a). Single amino acid additions to proteins in eukaryotic and bacterial flagella can be visible or invisible to natural selection depending on the copy number of the protein within the flagellum and cell (86) (Figure 4b). A comparison of energetic costs of DNA and Ne in viruses reveals that small stretches of DNA can persist in larger viruses even in the absence of a selective benefit (53).
Traits do not just burden the cell. They also endow it with benefits. It is the balance between costs and benefits that determines the net fitness effect—and that, together with genetic drift and mutation pressure, decides the fate of a trait and its elaborations. In many cases, it is difficult to quantify the fitness effects of trait variants, as the benefits often depend on unknown aspects of the external environment, but analyses of costs and benefits have been performed for vacuoles, osmoprotection, flagella, vesicles, and other traits, which we discuss in the following paragraphs.
Vacuoles in photosynthetic cells can be used to increase the average photon capture efficiency of chromophores by spreading these chromophores over a larger area with respect to the incident sunlight. This is the result of a central vacuole turning the cell into a shell of cytoplasm surrounded by a cell wall. At a fixed cytoplasmic volume (excluding the vacuole) the addition of a vacuole increases plasma membrane and cell wall costs and introduces vacuolar membrane costs and costs of maintaining vacuole osmotic balance. Comparing a spherical 5-μm-radius nonvacuolate cell to a spherical 11.1-μm-radius vacuolate cell with the same cytoplasmic volume shows that there is a 2.1-fold increase in cellular cost, from 4.2 × 1013 to 8.7 × 1013 ATP (mostly cell wall) (79). The benefit of vacuolation, by increased photon uptake, can exceed the cost if the chromophore concentration is higher than some threshold value (79; see also 80).
Cells in freshwater face a constant influx of water by osmosis, a problem that can be solved either by synthesizing a rigid cell wall or by constantly expelling water by contractile vacuoles. For a spherical algal cell with a radius of 5 μm and a cell division time of 24 h, the energetic cost of a cell wall is 2.75 × 107 ATP s–1, and the cost of operating a contractile vacuole is 3.78 × 107 ATP s–1, the cell wall being increasingly favored as the cell grows more slowly (75). The cost of the contractile vacuole is so high that there is a substantial selective pressure on keeping the cytoplasmic osmolarity low in organisms with lifestyles that exclude a cell wall, with large consequences for the operation of other cellular traits (75). In ciliates, the construction cost of a contractile vacuole is 1.8 × 105 ATP s–1 (assuming a cell division time of 24 h), which is only a minor contribution compared to the operation cost mentioned above (50). Cell envelope costs for bacteria have also been estimated (51, 95), typically making a major contribution to the cell budget (Figure 2f).
The flagellar construction cost in terms of both energy and nitrogen composition has been examined in a model dinoflagellate (83). In a scenario of this dinoflagellate swimming up in the water column to increase its light-harvesting capacity and down to increase its nitrogen uptake, the flagellum pays back for itself many times over (83). A similar analysis can be performed for cells of different volumes and swimming speeds that live in a homogeneous distribution of small-molecule nutrients. Only large cells can improve their nutrient uptake beyond the investment in flagella (86) (Figure 4c). A cost-benefit analysis can also be used to compare the capacities of the bacterial and eukaryotic flagella, which function by wildly different mechanisms. Plotting the swimming speed per ATP invested in flagella against cell volume reveals that the bacterial and eukaryotic flagella follow a common trend. They appear to be equally effective at generating swimming speed, though they function at different cell volume ranges (86) (Figure 4d). The cost-benefit relation for chemotaxis has also been examined (67).
Other cost-benefit analyses are of photon absorption (26, 77), active water transport and osmotic compound synthesis for the purpose of buoyancy (44, 82), the cell volume gain accomplished by taking up a proto-mitochondrial respiring symbiont (88), gene regulation in a variable environment (18, 38), and responses to high photon flux (84).
The costs and benefits discussed so far consist of systems as they exist in present-day species. For understanding evolutionary transitions or for understanding why seemingly plausible alternatives do not exist, it is important to perform cost-benefit analyses on cellular traits that deviate from what presently exists. Such analyses require models in which the investment in a trait can be varied and in which the benefit to the cell can be computed from this investment.
In the study of early endomembrane evolution, direct examples of intermediate states are lacking and need to be constructed from plausible subcellular parts. One example of such a cellular part is the vesicle, which has been extensively studied in contemporary species (e.g., 89). Vesicle properties, such as size, plasma membrane area occupancy, construction cost, and construction time, can be combined in a model with functional elements that reside on those vesicles. Placing nutrient transporters on those vesicles yields a primitive form of pinocytosis, while adding Sec translocases facilitates membrane protein insertion and yields a proto-endoplasmic reticulum. Cost-benefit analyses can be performed by varying the abundance (investment) of those vesicles, revealing that pinocytosis of small-molecule nutrients does not improve fitness under sensible parameter ranges but that the proto-endoplasmic reticulum does improve fitness (87).
Cellular traits exhibit both fitness costs and benefits, with their balance determining the likelihood of establishment and continued existence of a trait in a population. The relative energetic cost of a cellular trait confers an inescapable baseline fitness penalty. An absolute energetic cost of a cellular trait is assembled by combining the known costs of basic building blocks such as amino acids, lipids, sugars, and nucleotides. This is converted into a relative energy cost by dividing it by the cell energy budget. Absolute costs have been estimated for a circadian clock, cell walls, flagella, vacuoles, osmoprotection, mitochondrial and nuclear membranes, and NPC, among others. Cell budgets have been estimated for species across the tree of life. Combining the absolute trait costs and the cell budgets helps reveal what evolutionary scenarios are plausible and what cellular configurations are possible.